Saving observation states and actions at the end of each episode

[dhruvkm2402]

Hello,
I've implemented the tool and it's really good w.r.t flexibility and one code file which is so much easier. Thank you for that.
I wanted to know if we can save the states and actions at the end of each episode?, as that would be a useful data later.

[Toni-SM]

Hi @dhruvkm2402

It is possible to save, in persistent storage, all environment transitions (state, action, next state, reward, terminated, and truncated) and any other information stored in memory automatically when the memory is filled

Supported formats are PyTorch (.pt), NumPy (.np), and Comma-separated values (.csv)

Please visit the Memory export/import section

[dhruvkm2402]

Thank you for pointing it out